Predicting fire at global scale: Preliminary steps for a Spatial-temporal Analysis of Global Fire Time Series

Laboratory of Geo-Information Science and Remote Sensing
  Education
  Research
  Publications
  Models
  News & Calendar
  News
  Calendar
  Archive
  News
  Calendar
  2011
  2010
  2009
  2008
  2007
  2006
  2005
  2012
  Staff
  Equipment
  Contact details
  Workshops

23 Oct 2008 14:00 - 23 Oct 2008 14:30
Unit: Laboratory of Geo-Information Science and Remote Sensing
Location: Atlas 2
Organisation: Wageningen University

 By Xiaoyu Guan: 

Summary
This research is focused on explaining and predicting fire occurrence at the global scale using environmental variables. To understand the role of fire in ecosystems and to explain fire occurrence, fire characteristics (fire frequency, return intervals and fire duration) were related to environmental variables (e.g. climate, precipitation, vegetation biomass, elevation and soil data). Weekly synthesis of global burned scare maps (Global Burned Surface Product – GBS) which covered 17 years from 1982 to 1999 (excluding 1994) with a pixel resolution of 8km were used in the study. Each cell of each week image contains 0, which means a cell without fire, or 1, which means the week with fire. The other data sources were collected from several organizations with different formats. Correlation test and multiple regression models were used to analyse the relation between the environmental variables and fire characteristics at global scale. The expected relations were 1). Fire occurrence is expected to have a unimodal relation with precipitation. 2). Temperature correlates positively with fire occurrence. 3). Normalized Difference Vegetation Index (NDVI) correlates positively with fire occurrence. 4). Soil variables have a positive correlation with fire occurrence. 5). Elevation has a positive correlation with fire occurrence. The results shows that mean annual temperature and mean annual NDVI have a significant correlation with most of fire variables. Conversely, the other variables do not have a significant correlation with fire variables. The frequency of explaining factor in a series of models has been counted in a bootstrap procedure. Two linear models could be formulated with the explaining factors which have high frequency. One model’s adjusted R-square is 0.327, another model’s adjusted R-square is 0.398. The mean annual temperature and mean annual NDVI are more important than other factors in explaining fire frequency at global scale. Furthermore, in linear model, a unimodal relation is found between NDVI and fire frequency. Soil: percentage of clay is positively correlate with fire variables. however, the significant level is relative low. This study conducts that a linear relation is between environmental variables and fire variable at global scale.
 
Print this activity